Department of Economics

This paper presents the first laboratory study of risk-sharing without commitment. Our experiment captures the main features of a simple model of voluntary insurance between two agents. In the model, two individuals interact over a potential infinite horizon and suffer random income shocks. Risk-averse individuals have incentives to smooth consumption by making transfers to each other. These transfers being voluntary, only self-enforcing risk-sharing arrangements are possible: transfers can never be so large as to tempt individuals to renege on them. This constraint, when binding, has strong implications for the shape of the constrained optimal risk-sharing arrangement. In our experiment, participants are matched in pairs. Each period, one of them, randomly drawn, receives a given amount in addition to its regular income. After observing both incomes, each person in a pair chooses a non-negative transfer to make to the other person. Two features of the experimental design are crucial. First, it is common information that all pairs will be dissolved at the end of each period with a given probability. Participants are informed when this occurs and randomly re-matched. This replicates the effect of infinite-horizon and discounting in the model. Second, at the end of the experiment, a single period is randomly drawn to count for cash payment. This feature is essential for individuals to care about the utility outcome of each period. We find evidence generally consistent with risk sharing, with most transfers coming from individuals who received h in the period. Moreover, in support of the theory, transfers are much higher with a higher continuation probability and they also are highly correlated with the individual’s degree of risk aversion. However, while the model predicts an increase in transfers with ex ante inequality, we observe the opposite effect. This may reflect considerations of identity or group membership.